The 2026 World Cup kicked off today at Estadio Azteca with Mexico vs. South Africa, and with the opening whistle the largest mass deployment of artificial intelligence in Latin America’s history was switched on. Google launched its Gemini features for the tournament —real-time scores on the lock screen, AI-generated sports summaries and photo templates with national team jerseys— and also confirmed that Argentina’s national team uses Gemini for tactical analysis and injury prevention. With Android on more than 80% of devices in the region, the World Cup turns hundreds of millions of Latin American phones into de facto Gemini terminals, without most users actively choosing it.
But the same country hosting the opening match also deployed “Nexus,” a facial recognition and vehicle tracking platform for the event’s logistics hubs, with no announced date for its withdrawal. These are two layers of the same tournament: consumer AI, visible and friendly on the phone screen, and surveillance AI, invisible at the stadium gate. Google describes its features as “a tool to connect fans with the game they love”; state facial recognition runs in parallel, without that friendly narrative and without a timetable for being switched off. For the region, the World Cup thus works as an enormous laboratory where the two faces of the same technology are tested at the same time and on the same crowd.
Also today
- SpaceX sets its IPO price at $135 per share — It debuts tomorrow on the Nasdaq in what would be the largest public offering in history ($1.75 trillion). The prospectus revealed a data point that connects with the big AI labs: Anthropic pays about $1.25 billion a month for the Colossus compute clusters. It is the first stock market instrument directly tied to frontier AI compute that is accessible from the region.
- Claude comes natively to the iPhone — A new unified protocol in iOS 27 makes it possible to swap between Apple’s on-device model, Gemini and Claude without rewriting code; more complex analysis tasks would be routed to Claude. Available to developers today; for end users with iOS 27, in the fall.
- A US federal court examines whether AI agents can “visit” websites — The Ninth Circuit Court of Appeals heard arguments today in Amazon v. Perplexity, the first federal appellate precedent on whether an autonomous agent counts as an authorized visitor. Digital rights organizations such as the ACLU and the EFF are backing Perplexity. If Amazon wins, the web agent ecosystem in the region falls into a legal gray zone.
- OpenAI integrates its models into Oracle Cloud — Oracle customers will be able to use their existing credits to access OpenAI’s models without opening a separate account. Oracle has an enterprise presence in Brazil, Mexico and Chile.
- Gemini went down for eight hours just before the World Cup — A global outage affected Gemini on June 10 for much of the day. Google promised Gemini would be at the center of the tournament; on the eve, it was not there.
In the region
On June 10 the European Union published its final Code of Practice on labeling AI-generated content, the first operational technical standard —based on the C2PA standard, which adds a mark of origin to files— that requires deepfakes in content of public interest to be identified. It takes effect on August 2. The underlying pattern is that three very different AI governance models now coexist —the voluntary approach of the United States, the European obligations and this new labeling code— and none was designed with Latin American participation or is binding on its territory. Colombia is the most exposed case: it reaches the June 21 presidential runoff without an equivalent rule in force.
Launches
- Gemini for the 2026 World Cup — Daily sports summaries, lock-screen scores and photo templates with national team jerseys. Free in the Gemini app for Android and iOS.
- Claude for Apple’s framework (iOS 27) — A Swift package that integrates Claude into Apple’s models framework, with the option of combining it with the device’s local model. Available to developers today; for the public with iOS 27 later in the year.
Threads we’re following
For weeks we have been following Colombia’s presidential election as the acid test of AI disinformation in the region. Today another chapter was added: ten days before the runoff, the National Registrar warned about the violence that could result from AI-generated disinformation. There are already four documented deepfakes, and cloned audio is available for about $10 a month, while the Colombian law that should punish them still has no effective sanctions. It is the biggest regulatory asymmetry in electoral AI we have seen in the region: the technology to fake is cheap and abundant; the rule to stop it still has no teeth.
If the same technology can be, in the same place and at the same time, a toy that brings us closer to the game and a surveillance infrastructure that watches us, who decides which of the two layers stays on when the tournament ends?
About this entry. It is generated automatically from public sources, without human review before publication. It may contain errors of interpretation or summary; please check each story against its original source (the links lead there) before citing it or making decisions based on it.
Doble Click is written with Anthropic models.